The First 90 Days: Turning Opened Accounts into Primary Relationships

"Accounts opened" is the wrong number

Most growth reports count accounts opened. The board sees the number rise, marketing takes credit, and the acquisition budget gets renewed. Nobody asks how many of those accounts hold a balance ninety days later.

That question matters because of a behavior J.D. Power describes as quietly making a new account primary. In its Q3 2025 data (October 2025), 72% of additional or replacement checking accounts were opened with a different institution than the one the customer already used, and 54% of those became the primary account. The old institution did not lose a customer in any way its systems would notice. The account stayed open. The balance drifted down. Direct deposit moved.

Rivel's data, from 227,173 interviews in Q1 2026 (via The Financial Brand, May 2026), shows the same pattern from the other side: 41% of Millennials with more than $100,000 have previously left an institution without closing the account. Your attrition dashboard counts closures. A dormant account never registers as one.

The movement is accelerating. J.D. Power's 2026 U.S. Retail Banking Satisfaction Study (March 2026) found 20% of customers moved money away from their primary bank in the past three months, up from 17% the prior year, and the average customer now holds three deposit accounts across institutions. Curinos (Q1 2026 review, March 2026) found 50% of switchers had four or more checking relationships in 2024, against 7% in 2019, and fintechs now capture nearly 40% of new banking relationships.

Then there is the cost side. The Financial Brand (August 2026) puts fully loaded acquisition cost at about $489 per checking account, and above $1,000 at some institutions. Paying that for an account that becomes someone else's primary within twelve months is not growth. It is a row in the core and a cost on the books.

The four events that decide primacy

Primacy is not a feeling. It shows up in the core as four events, and each has a measurable drop-off.

A funded balance. This is the first and largest leak. Cornerstone Advisors and Alkami's 2026 Digital Banking Performance Metrics (May 2026) count 3.36 digital checking applications abandoned per completed account, roughly 9,000 lost accounts per institution. Funding is one of the steps where that loss happens, and the method matters: MX reports drop-off as high as 49% with micro-deposits and as low as 1% with instant account verification.

A direct deposit that lands. Direct deposit decides which of a person's three accounts is the primary one, because payroll is where the money enters. It is also the step institutions most often leave for a follow-up email, and that email competes with every other institution where the accountholder already banks.

A joint owner or household added. An account with two owners, or one linked to a spouse's or child's account, is harder to leave. Most flows cannot complete it: the second person needs their own identity verification and watchlist screening, so platforms push it to a branch visit or a PDF. Among Gen Z, per Adrenaline research in The Financial Brand (April 2026), 60% already use multiple providers and 20% expect to switch primary within six months.

A first digital session that leads to a second product. The Cornerstone and Alkami report found engaged digital users hold 1.56 more products than others and log in about 17 times a month; 87% of checking accounts now have an active digital user. If the first login does not happen in the first thirty days, the second product rarely happens in the first ninety.

Step 1 — Do it in the session

The applicant is never more willing to complete an activation event than in the minutes after approval. They have their ID out, their phone in hand and their attention on your institution. Every day between that moment and the follow-up campaign is a day the account sits at zero.

Most institutions treat onboarding after account opening as a campaign problem. It is a sequencing problem. On Cotribute, funding, joint owner, disclosures and the direct-deposit switch all complete before the tab closes, and each writes to the core in real time rather than in a nightly batch.

Funding runs through instant account verification, so the applicant links an external account and moves an opening balance in the same sitting instead of waiting on micro-deposits. A one-time passcode gates the money movement.

A joint owner is added inside the flow with their own identity verification and watchlist screening, so the second signer is a screened person on the core record at day one, not a note for the branch. Disclosures are versioned and signed electronically in the session, so the account is fully documented at the moment it is created.

The direct-deposit switch runs in the same session. Cotribute orchestrates it; the switch itself runs on a provider your institution contracts with directly, on your contract and your pricing. Integrated providers include Atomic, ClickSWITCH, InstaSwitch, Knot, Pinwheel and Plaid, among others. Cotribute charges no connector fees and does not build or resell the switch.

The account is therefore funded, owned, documented and receiving payroll when it is created, not thirty days later. Across the Cotribute portfolio, the median institution sees 58 of every 100 applications reach a funded account, against an industry figure of 23 of 100 (Cornerstone Advisors and Alkami, May 2026). Abandonment runs at 42% against 77%.

Capitol Credit Union of Texas ($225M) is the clearest case. CEO Pierre Cardenas: "Our initial results have been phenomenal — on average, $10,000 in new deposits per account opened and 70% reduction in processing time." Digital became the credit union's top source of new accounts within 90 days.

None of this requires a core conversion, an LOS replacement or a digital banking replacement. Cotribute is real-time on 11 of the 13 cores it integrates, including all six Fiserv cores, Jack Henry Symitar and Corelation KeyStone. The core stays the system of record.

Step 2 — The 90-day scoreboard

Once activation happens inside the application, you can measure it. Replace "accounts opened" with six numbers, reported monthly by cohort, and take them to the same meeting where the acquisition number used to go.

Three notes on reading it.

First, report by monthly cohort, not as a rolling average. A cohort view shows whether the change you made in March moved the June numbers; a rolling average hides it.

Second, the last row tells you the truth. Every institution has zero-balance accounts. The question is whether the share is falling as you move activation into the session. If it is not, the leak is upstream of your outreach, and no campaign will fix it.

Third, the funded rate at day one is the only row with a public benchmark: 23 of 100 across the industry, 58 of 100 at the Cotribute median, 85 of 100 at the best-in-class client. The other five rows have no reliable industry number, so establish your own baseline before you change anything.

When the CFO asks what the scoreboard is worth, keep deposits gathered (balance sheet) and operating cost released (income statement) on separate lines. They are never summed into a single ROI figure.

Step 3 — Next-best-step, human approved

The scoreboard tells you which cohorts are stalling. It does not tell your staff which twelve accounts to call this morning. That is the job of a relationship growth agent.

Cotribute's AI Growth Agents have been in production since June 2025. There are three: acquisition, cross-sell and relationship growth. The relationship growth agent reads the same events the scoreboard reads, from the origination platform and the core, and produces a recommended next step for each account: a member who funded but has no direct deposit at day 20, a customer whose joint applicant verified but never logged in, a household with a certificate maturing next month and a savings account still at its opening balance.

Each recommendation goes to a person. Staff, or the rules your institution has configured, approve and act. The agent does not send the message, move the money or open the product. It does not approve, decline or refer anything. No accountholder data enters a public model.

The reason to keep approval with staff is practical. The 1.56 additional products that engaged digital users hold (Cornerstone Advisors and Alkami, May 2026) describe a relationship, and relationships are won by the person at the branch or on the phone who knows the account. The agent finds the accounts; the conversation is still yours.

The approval chain is also auditable. Every configuration change is logged, with what changed, who changed it and when, so the rule set can be reconstructed as it stood on any date.

What to ask any vendor

Five questions separate a platform that does activation from one that talks about it.

Does funding write to the core in-session or in a nightly batch? If the balance appears tomorrow, the applicant's first digital login shows $0, and your day-one funded rate is measured against the wrong day.

Whose contract does the deposit switch run on? The right answer is yours. A vendor that resells the switch is taking a margin on your own payroll traffic. Ask whether the platform orchestrates a provider you contract with directly, and whether it charges connector fees.

Can a joint owner complete identity verification and watchlist screening inside the applicant's session? Most platforms accept a second name and leave verification to the branch. That makes the joint owner a stipulation rather than an owner.

Does the platform recognize an existing accountholder? An existing member or customer opening a second product should be pre-filled from the core, skip the steps already satisfied and not be re-verified. If the platform treats them as a stranger, the second-product row on your scoreboard stays flat.

Can I get the four activation events out as data? You need funded, direct deposit, joint owner and first login as timestamped events, joined to the core record, so the scoreboard runs on your data and not on the vendor's report. Ask how origination data is exposed, who can query it and whether every query is audited.

Frequently asked questions